Gene Therapy (1999) 6 , 865–872  1999 Stockton Press All rights reserved 0969-7128/99 $12.00 http:/ / www.stockton-press.co.uk/ gt Cytotoxic T lymphocyte response against non- immunoselected tumor antigens predicts the outcome of gene therapy with IL-12-transduced tumor cell vaccine M Rodolfo, C Zilocchi, B Cappetti, G Parmiani, C Melani and MP Colombo Experimental Oncology D, Istituto Nazionale per lo Studio e la Cura dei Tumori, Milan, Italy The colon adenocarcinoma C26, carrying two endogenous related tumor antigens, whereas nonresponders had CTL tumor-associated antigens (TAA) recognized by CTL, has that recognized preferentially the FR antigen. CD8 from been transduced with the gene coding for the human folate responder mice were characterized to release high levels receptor  (FR) as an additional antigen in order to study of granulocyte–macrophage (GM)-CSF upon antigen the efficacy of vaccination against a tumor expressing mul- stimulation. Tumors obtained from mice that died despite tiple antigens. A dicistronic vector was used to transduce vaccination lost expression of the FR transgene but main- the IL-12 genes to create C26/IL-12/FR that has been tained expression of endogenous C26 antigens. Immuno- used as a cellular vaccine to treat mice bearing lung met- selection against FR antigen was not observed in tumors astases of C26/FR. After vaccination mice were partially from non-vaccinated controls and from CD8-depleted vac- splenectomized and splenic lymphocytes frozen and used cinated mice. Down-regulation of FR antigen expression retrospectively to study in vitro CD8 T cell response related was due, at least in part, to methylation of retroviral vector to the treatment outcome. Vaccination cured 50% of mice long terminal repeat promoter since FR expression was and the effect was CD8 T cell dependent. Mice either cured partially restored, ex vivo, by treatment with 5-aza-2'- (responders) or not cured (nonresponders) by vaccination deoxy-cytidine (aza). These results indicate that CD8 T developed tumor-specific CTL. However, analysis of CTL cell-mediated immunoselection and production of GM-CSF specificity and pCTL frequencies revealed that responders are determining factors for the efficacy of tumor vaccines. had a predominant CTL activity against endogenous C26- Keywords: cytokine gene transfer; tumor immunity; CTL Introduction Active tumor immunotherapy by vaccination with irradiated neoplastic cells is expected to activate an immune response against the entire repertoire of tumor- associated antigens (TAA), a few of which have been molecularly defined. 1,2 Data have emerged that encour- age the use of cellular-based vaccines for melanoma patients. 3,4 Active immunotherapy can be further improved by genetic modification of tumor cells through insertion and expression of cytokine genes. 5 In fact, pre- clinical studies have shown that the injection of such engineered tumor cells induces a local inflammatory reaction leading to the activation of a systemic antitumor memory response and, in some cases, to the cure of estab- lished tumors. 6 A major issue in the use of cell-based vaccines is the role of the different antigens or subsets of antigens expressed by tumor cells in eliciting a T cell response determining tumor regression. In fact, CTL responses Correspondence: M Rodolfo, Experimental Oncology D, Istituto Nazionale Tumori, via Venezian, 1-20133 Milan, Italy Received 12 August 1998; accepted 25 November 1998 against immunodominant antigens, such as melanoma- associated MART-1/MelanA or Pmel17/gp100, can mediate in vivo destruction of antigen-positive tumor cells but also select for antigen loss variants. It has been reported that during tumor progression, antigenic vari- ants can appear which may loose not only the HLA class I molecules, but also T cell epitopes. 7–9 In this study we have analyzed the CD8 T cell response associated with the therapeutic activity following IL-12- transduced tumor cell vaccination. Considering that immunization may direct the immune response against an antigen that is therapeutically irrelevant or lost because of immune selection, we have set up a model system in which vaccination therapy is directed to a tumor carrying more than one known antigen. This enables testing which antigen is the relevant target of host-induced immune response but also the relationship between immunodominance and immunoselection. We used the C26 colon carcinoma which expresses as immu- nodominant TAA the L d restricted peptide gp70 (423– 431), called AH1, encoded by env gene of the endogenous ecotropic murine leukemia virus (MuLV) emv-1. 10 In addition, C26 express at least one additional, but as yet molecularly undefined, subdominant tumor rejection antigen recognized by a CTL clone, 11 TA9A (tumor anti-